Fast aberration correction in 3D transcranial photoacoustic computed tomography via a learning-based image...

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Computational Science

Background:

• Transcranial photoacoustic computed tomography (PACT) shows promise for neuroimaging.
• Skull-induced aberrations significantly challenge image reconstruction quality in transcranial PACT.
• Existing optimization-based reconstruction methods are computationally intensive and sensitive to skull property accuracy.

Purpose of the Study:

• To investigate a novel learning-based image reconstruction method for 3D transcranial PACT.
• To assess the performance and robustness of this method against traditional approaches.

Main Methods:

• A deep learning-based image reconstruction algorithm was developed for 3D transcranial PACT.
• The method was validated using numerical simulations with stochastic 3D head phantoms.
• Experimental data from a physical human skull phantom were used for further assessment.

Main Results:

• The learning-based method successfully produced accurate 3D reconstructed images.
• The approach demonstrated robustness to inaccuracies in estimated skull properties.
• Computational time was substantially reduced compared to optimization-based reconstruction methods.

Conclusions:

• A learned image reconstruction method offers a viable and efficient solution for 3D transcranial PACT.
• This technique overcomes key limitations of current methods, enhancing neuroimaging potential.
• This represents the first reported demonstration of a learned reconstruction method for 3D transcranial PACT.